Traumascan : assessing penetrating injury with abductive and geometric reasoning / Omolola Ijeoma Ogunyemi.

Format:

Book

Manuscript

Microformat

Thesis/Dissertation

Author/Creator:

Ogunyemi, Omolola Ijeoma.

Contributor:

Webber, Bonnie Lynn, advisor.

University of Pennsylvania.

Language:

English

Subjects (All):

Penn dissertations--Computer and information science.

Computer and information science--Penn dissertations.

Local Subjects:

Penn dissertations--Computer and information science.

Computer and information science--Penn dissertations.

Physical Description:

xiii, 140 pages : illustrations ; 29 cm

Production:

1999.

Summary:

To perform computer-aided assessment of the injuries that may result from penetrating trauma, the relationship between human anatomy, physiology, and physical manifestations of injury must be modeled. Since penetrating injury involves actual physical damage to regions of the anatomy, any effective method for assessing penetrating injury must take into consideration the mechanism that produces the injury. However, it may not be possible to accurately capture the extent of damage associated with a mechanism of injury, and there may be only partial information available about patient manifestations of injury. These two problems introduce uncertainty into the assessment process. Assessment of ballistic injuries is further complicated by the fact that many different entry-to-exit wound pairings are possible when multiple external wounds exist, and the number of possible pairings increases exponentially with the number of external wounds.

We address these and other problems in our prototype system, TraumaSCAN, which assesses the effects of penetrating trauma by combining geometric reasoning about potentially injured anatomic structures with abductive reasoning about the consequences of these injuries. Geometric reasoning in TraumaSCAN makes use of 3D polygonal surface representations of the anatomy and 3D polygonal surface models of the damage associated with a mechanism of injury. Abductive reasoning is performed with Bayesian networks, which allow us to develop probabilistic causal models of the relationships between anatomic injuries, diseases, and patient findings. Combining geometric and abductive reasoning ensures that knowledge of anatomy, physiology, and manifestations of injury are integrated in the assessment process. Our approach uses probabilities as a means of mediating between the different forms of reasoning, provides a means of initial diagnosis that incorporates the mechanism of injury in a significant way, allows visual exploration of the potential consequences of penetrating injury which complements written feedback, and enables the systematic examination of alternative injury outcomes. We present results obtained from testing TraumaSCAN on a number of actual penetrating trauma cases for which the correct diagnoses are known.

Notes:

Supervisor: Bonnie Lynn Webber.

Thesis (Ph.D. in Computer and Information Science) -- University of Pennsylvania, 1999.

Includes bibliographical references.

Local Notes:

University Microfilms order no.: 99-26180.

OCLC:

244970670